#hal / OpenAI paid Kenyan workers less than $2 per hour to make ChatGPT less toxic.

From Time magazine:

The work was vital for OpenAI. ChatGPT’s predecessor, GPT-3, had already shown an impressive ability to string sentences together. But it was a difficult sell, as the app was also prone to blurting out violent, sexist and racist remarks. This is because the AI had been trained on hundreds of billions of words scraped from the internet—a vast repository of human language. That huge training dataset was the reason for GPT-3’s impressive linguistic capabilities, but was also perhaps its biggest curse. Since parts of the internet are replete with toxicity and bias, there was no easy way of purging those sections of the training data. Even a team of hundreds of humans would have taken decades to trawl through the enormous dataset manually.

By “toxic and bias” what they really mean is content that “described situations in graphic detail like child sexual abuse, bestiality, murder, suicide, torture, self harm, and incest.”

Perhaps $2 per hour is a lot in Kenya, but that’s no excuse for exploiting people. And given the absolute cesspool that OpenAI asked these workers to sift through, they should have included additional hazard pay.

#hal / ChatGPT Will Disrupt McKinsey Before it Disrupts Google.

Greg Larkin shares an interesting insight on the future of companies like McKinsey & Company, Boston Consulting Group (BCG), and Bain & Company. Or as he calls them, the “Term Paper Industry.

ChatGPT will emerge as a viable first draft for people who need better answers than what the Term Paper Industry historically offers. I believe that when they seek an outside perspective for a complex problem they will build a baseline hypothesis of the current state using Chat GPT, and will then pressure test the results by seeking out the actual ‘experts’ whose insights stem from years of hard-earned trial and error. This will lead to defections from the term paper companies. These defections will first be small, but will then become large and irreversible, and eventually lead to large scale fragmentation in management consulting, and investment banking.

Using AI to generate a hypothesis of the current state is intriguing, especially if it’s paired with design labeled (I’m totally rolling my eyes as I type this) practices like “design research” and “design thinking.”

#hal / Something in the Way: AI-Generated Images and the Real Killer.

A moving perspective from a successful cover illustrator:

Machine-driven narrative will get better and better as the taste for popular narrative gets more and more watered down. And again, it won’t necessarily be because the AI does better work than working visual artists and writers, but because the audience settles for accepting the mass convenience of “good enough,” drowning out the need for quality of content. Audiences cherish convenience over quality. It’s what drives our ethos, at least here in the U.S. No reason to believe people are suddenly going to wake up and change.

The arrival of AI generated-images (and how industry chooses to use it) goes far beyond what happened upon the advent of photography or even Photoshop. This is NOT the same conversation. AI is a tidal shift from the center of human context that defines meaning. We’re not in Kansas anymore, Toto. Choose your yellow brick road very wisely, folks.

“Good enough” is the perfect description for AI content. Paired with the Internet, you have the right combination to placate a majority of the population—either by personal choice or class discrimination. Which means AI is the Walmartification of the Internet.


The machines will destroy us with personalized generated content to amplify distrust.

I’m not a fan of artificial intelligence. It’s not that I don’t see the positive impact of these technologies. I’ve worked on AI based projects at IBM that were designed to perform tedious bureaucratic work quickly. My concern is that we’re making very powerful yet easy-to-use tools to generate deceptive written and visual content (audio and video aren’t too far off) that is becoming more and more difficult to interpret as authentically created by a human. We’re on a path to destroy the Internet—and maybe worse—as we know it today. I’ve been feeling like an old man yelling at a cloud on this idea, and then I read The Expanding Dark Forest and Generative AI by Maggie Appleton.

The dark forest theory of the web points to the increasingly life-like but life-less state of being online.

Most open and publicly available spaces on the web are overrun with bots, advertisers, trolls, data scrapers, clickbait, keyword-stuffing “content creators,” and algorithmically manipulated junk.

To complicate matters, language models are not the only mimicry machines gathering speed right now. Image generators like Midjourney, DALL-E, and Stable Diffusion have been on a year-long sprint. In January they could barely render a low-resolution, disfigured human face. By the autumn they reliably produced images indistinguishable from the work of human photographers and illustrators.

There’s a swirl of optimism around how these models will save us from a suite of boring busywork: writing formal emails, internal memos, technical documentation, marketing copy, product announcement, advertisements, cover letters, and even negotiating with medical insurance companies.

But we’ll also need to reckon with the trade-offs of making insta-paragraphs and 1-click cover images. These new models are poised to flood the web with generic, generated content.

If this doesn’t unsettle you to the point of needing to shift in your chair—it should. And if that is the case then either you embrace and fan the flames of anarchy or you have not had to contend with friends and family members who are easily baited by convincing content created by trolls (human trolls ) on the internet. These AI tools are going to make it a thousand times worse.

And yet, despite all of the problems trolls have created for societies around the world in the past seven years, we are willingly and eagerly creating tools to generate even content that will look and feel even more convincing—Vast amounts of garbage content—by providing just a few instructions and a click of a button. No good can come from this.

Science fiction has traditionally portrayed artificial intelligence as an entity that becomes self-aware, determines mankind is a threat, and takes over military resources or builds a robot army to destroy humans. I think they got it wrong because all you need is for a robot to generate content that “sounds” human and broadcast it on the internet. We’ll end up killing each other without the AI having to launch or build a single thing. And there are many people in the technology industry pulling late shifts to make these tools even more powerful, easy to use, and easy to access. Today people are kicking the tires of AI to generate fictional works of art all in the name of fun or curiosity, but it won’t be long before someone uses these tools for evil. And the machines will be watching and learning.

The internet is already littered with so much crap—these tools will flood our digital spaces with useless, generated content. And as Maggie suggests, we are not prepared for this outcome.

Many people will say we already live in this reality. We’ve already become skilled at sifting through unhelpful piles of “optimised content” designed to gather clicks and advertising impressions.

4chan proposed dead internet theory years ago: that most of the internet is “empty and devoid of people” and has been taken over by artificial intelligence. A milder version of this theory is simply that we’re overrun with bots. Most of us take that for granted at this point.

But I think the sheer volume and scale of what’s coming will be meaningfully different. And I think we’re unprepared. Or at least, I am.

No, Maggie, you’re right. The world is so, so unprepared. Idiocracy, here we come!